STDP-Driven Automated Retinal Circuit With 7-nm FinFET for Motion and Looming Detection: A Hybrid Model With Image Analysis

This paper presents a spike-timing-dependent plasticity (STDP) driven silicon retina designed using 7nm FinFET technology for detecting object motion and looming threats inspired by the biological visual system. The proposed neuromorphic architecture integrates an automated motion spiking circuit with STDP and advanced image analysis, mimicking natural retinal processing to identify potential hazards. Two specialized circuits are introduced to generate bipolar spiking signals for object motion and looming detection. These spikes are processed by an STDP-based spiking neural network, qualifying real-time distance estimation of approaching objects. Model Results demonstrate accurate collision time prognosis based on object dynamics. YOLOv5 and MiDaS enhance object recognition and depth perception, improving situational awareness. The proposed model uses a hybrid strategy combining object detection and depth estimation to identify objects with different spatial distances and confidence levels accurately. Experimental results show a motion detection accuracy of 93.2% and an average latency of around 15 ms, 66 frames per second, which proves this is suitable for time-sensitive applications. Because the 7nm FinFET technology improves system efficiency and responsiveness, it can be used in real-time applications where high-precision motion detection is necessary in dynamic conditions. However, potential challenges include scalability for large-scale deployment and hardware constraints associated with real-time processing on FinFET-based circuits. The proposed system can be used for robotic vision systems, autonomous vehicles, and other contexts where real-time visual processing and threat detection are needed.